Flow controlis used by the sender to limit their sending rate to what the receiver is capable of receiving. A receiver does not have unlimited buffer space to store segments waiting for the application to read them. As shown in the picture below, datagrams come in from the network and are stored in the buffer until delivered to the application. If the application reads at a slower rate than the datagrams arrive, this buffer will begin to fill. Once it is full, the datagrams will have to be dropped.

Clearly there is no point in the sender sending more data than the receiver has space to store. But how can the sender know how much buffer space the receiver has?

It's also important to note that both ends of a TCP connection can send data. For example a web server can return web pages over a connection, but the client can also send data through a POST over the same connection. So the client needs to know how much data the server is able to receive and the server needs to know how much data the client can receive.

Look carefully at the header fields in the TCP segments sent while transmitting the alice.txt file.

What header field is used by the sender of a segment to tell the other end how much buffer space it has left?

 http://gaia.cs.umass.edu/wireshark-labs/alice.txt

IP datagrams (currently) unused buffer space TCP data application (in buffer) process